It's true not yet implemented, but there is an idea to make a lang-file on the fly, i.e., connects the script that checks for all occurrences of Russian words (originally the interface is written in Russian) and as bypass creates a linguistic map, by analogy
$localize["Settings"]["UserName"]; in the file, for example eng.php, then you will only need to fill out as you work. Of course all this needs to be done on the UTF.
— no need to give up with the lang file and set variables in php files
— you can connect an unlimited number of languages
— decrease speed
Thoughts on the course
With speed you can play and create even a serialized array with reference to a specific php (all of the editing window, etc.), the set of language variables in these files is restricted to the standard set.
Language phrase will need to take in tag span id="lang" to increase speed.
PS While writing, I have read the above comments and realized that it is already in Zend_Translate stole my idea :)